2026
HiFiVe: High-Fidelity Vehicle Generation Leveraging Auto-Regressive 2D Generative Priors
Hongli Xiao*, Youjian Zhang*, Qi Zheng, Zhaohui Hu, Yaohui Jin, Xiaoguang Ren, Wenjing Yang, Long Lan (* equal contribution)
arXiv preprint 2026
This paper introduces HiFiVe, a training-free framework that enhances the texture and geometry of low-quality vehicle meshes by anchoring 2D generative priors to 3D geometric constraints.
HiFiVe: High-Fidelity Vehicle Generation Leveraging Auto-Regressive 2D Generative Priors
Hongli Xiao*, Youjian Zhang*, Qi Zheng, Zhaohui Hu, Yaohui Jin, Xiaoguang Ren, Wenjing Yang, Long Lan (* equal contribution)
arXiv preprint 2026
This paper introduces HiFiVe, a training-free framework that enhances the texture and geometry of low-quality vehicle meshes by anchoring 2D generative priors to 3D geometric constraints.
3DCarGen: Scalable 3D Car Generation via 3D-consistent Multi-view Synthesis
Hongli Xiao*, Youjian Zhang*, Yaohui Jin, Xiaoguang Ren, Wenjing Yang, Long Lan (* equal contribution)
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2026
This paper introduces 3DCarGen, a scalable single-view 3D vehicle generation framework that synthesizes 3D-consistent multi-view images from a single input. By combining an explicit 3D Gaussian Splatting prior with color-normal joint optimization, it successfully recovers high-fidelity and geometrically coherent 3D vehicle models suitable for autonomous driving simulation.
3DCarGen: Scalable 3D Car Generation via 3D-consistent Multi-view Synthesis
Hongli Xiao*, Youjian Zhang*, Yaohui Jin, Xiaoguang Ren, Wenjing Yang, Long Lan (* equal contribution)
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2026
This paper introduces 3DCarGen, a scalable single-view 3D vehicle generation framework that synthesizes 3D-consistent multi-view images from a single input. By combining an explicit 3D Gaussian Splatting prior with color-normal joint optimization, it successfully recovers high-fidelity and geometrically coherent 3D vehicle models suitable for autonomous driving simulation.
MM-TRELLIS: Point-Cloud Guided Multi-Modal 3D Vehicle Generation in Autonomous Driving
Hongli Xiao*, Youjian Zhang*, Yucai Bai, Chaoyue Wang, Yaohui Jin, Xiaoguang Ren, Wenjing Yang, Long Lan (* equal contribution)
IEEE International Conference on Robotics and Automation (ICRA) 2026
This work introduces a novel mesh-extraction method using LiDAR point clouds as diffusion guidance to correct geometry and scale inaccuracies in 3D generative models.
MM-TRELLIS: Point-Cloud Guided Multi-Modal 3D Vehicle Generation in Autonomous Driving
Hongli Xiao*, Youjian Zhang*, Yucai Bai, Chaoyue Wang, Yaohui Jin, Xiaoguang Ren, Wenjing Yang, Long Lan (* equal contribution)
IEEE International Conference on Robotics and Automation (ICRA) 2026
This work introduces a novel mesh-extraction method using LiDAR point clouds as diffusion guidance to correct geometry and scale inaccuracies in 3D generative models.
Skeleton2Stage: Reward-Guided Fine-Tuning for Physically Plausible Dance Generation
Jidong Jia*, Youjian Zhang*, Huan Fu, Dacheng Tao (* equal contribution)
arXiv preprint 2026
We leverage NVIDIA IsaacGym and imitation policies to fine-tune diffusion frameworks with specialized RL reward functions enforcing physical laws, significantly mitigating common artifacts like inter-penetration and foot sliding.
Skeleton2Stage: Reward-Guided Fine-Tuning for Physically Plausible Dance Generation
Jidong Jia*, Youjian Zhang*, Huan Fu, Dacheng Tao (* equal contribution)
arXiv preprint 2026
We leverage NVIDIA IsaacGym and imitation policies to fine-tune diffusion frameworks with specialized RL reward functions enforcing physical laws, significantly mitigating common artifacts like inter-penetration and foot sliding.
2025
D2GS: Dense Depth Regularization for LiDAR-free Urban Scene Reconstruction
Kejing Xia, Jidong Jia, Ke Jin, Yucai Bai, Li Sun, Dacheng Tao, Youjian Zhang# (# corresponding author)
Neural Information Processing Systems (NeurIPS) 2025
We resolve LiDAR projection inaccuracies in urban scene reconstruction via dense depth estimation and completion, leading to a LiDAR-free reconstruction pipeline.
D2GS: Dense Depth Regularization for LiDAR-free Urban Scene Reconstruction
Kejing Xia, Jidong Jia, Ke Jin, Yucai Bai, Li Sun, Dacheng Tao, Youjian Zhang# (# corresponding author)
Neural Information Processing Systems (NeurIPS) 2025
We resolve LiDAR projection inaccuracies in urban scene reconstruction via dense depth estimation and completion, leading to a LiDAR-free reconstruction pipeline.
2023
2022
Semantically-Consistent Dynamic Blurry Image Generation for Image Deblurring
Zhaohui Jing*, Youjian Zhang*, Chaoyue Wang, Daqing Liu, Yong Xia (* equal contribution)
ACM International Conference on Multimedia (MM) 2022
This paper introduces a novel method that generates semantic-aware global and local dynamic motion based on depth conditions to synthesize realistic motion-blurred images.
Semantically-Consistent Dynamic Blurry Image Generation for Image Deblurring
Zhaohui Jing*, Youjian Zhang*, Chaoyue Wang, Daqing Liu, Yong Xia (* equal contribution)
ACM International Conference on Multimedia (MM) 2022
This paper introduces a novel method that generates semantic-aware global and local dynamic motion based on depth conditions to synthesize realistic motion-blurred images.
2021
Exposure Trajectory Recovery from Motion Blur
Youjian Zhang, Chaoyue Wang, Stephen John Maybank, Dacheng Tao
IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) 2021
Performing image deblurring by recovering exposure trajectories directly from motion-blurred images.
Exposure Trajectory Recovery from Motion Blur
Youjian Zhang, Chaoyue Wang, Stephen John Maybank, Dacheng Tao
IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) 2021
Performing image deblurring by recovering exposure trajectories directly from motion-blurred images.
2020
Video Frame Interpolation without Temporal Priors
Youjian Zhang*, Chaoyue Wang*, Dacheng Tao (* equal contribution)
Neural Information Processing Systems (NeurIPS) 2020
Proposes a video frame interpolation framework that operates robustly with arbitrary exposure time and temporal intervals.
Video Frame Interpolation without Temporal Priors
Youjian Zhang*, Chaoyue Wang*, Dacheng Tao (* equal contribution)
Neural Information Processing Systems (NeurIPS) 2020
Proposes a video frame interpolation framework that operates robustly with arbitrary exposure time and temporal intervals.